Zhiyuan Zhang
Community@zhiyuan-zhang0206 · Los Angeles
Working in AI; MSc. of Stat, UCLA; Graduated from Tsinghua University
Agent Skills by Zhiyuan Zhang
Showing 97 vetted skills indexed across 1 GitHub repositories.
impersonator-guide
Operate an Ava impersonation lease using the Ava CLI and Python SDK under a borrowed identity.
ship-a-change
Ships repository changes through worktree, rebase, PR, CI, merge queue, and cleanup.
add-a-migration
Creates, applies, and rolls back timestamped SQL database migrations with paired down files.
deploy-ava-cluster
Installs Ava clusters and enrolls gateway or agent-runner machines across deployment shapes.
run-local-tests
Runs Python, frontend, and end-to-end test suites and diagnoses wedged local test infrastructure.
operating-ava-cluster
Diagnoses Ava cluster production alerts across disk, memory, connectivity, process, schedule, and delivery failures.
adversarial-review
Hunts correctness, security, contract, and fake-green defects in pull requests before merge.
measure-complexity
Measures cyclomatic complexity and maintainability index with radon to rank refactoring candidates.
write-a-pr-description
Writes PR descriptions with file-tree diffs, critical paths, and data flow notes.
ava-sweeper
Detects ten classes of technical debt in the Ava repository via greppable scans and reasoned audits.
inspect-a-trace
Correlate agent runs across Postgres checkpoints, Loki, Tempo, and Grafana to reconstruct failures.
worktree
Creates agent-named git worktrees for isolated coding changes before editing a repository.
pr
Validates commits, pushes, and pull request creation against repository hygiene and CI-readiness checklists.
testing
Runs pytest, ruff, pyright, vitest, eslint, and tsc checks after code changes.
conventions
Explains AGENTS.md auto-injection and shell-versus-files tool selection in Ava Code.
consolidation
Consolidates a git-based shared memory pool through commit, push, merge, and search index refresh.
reduce-context-switch-for-human
Routes delegated agent progress through manager roll-ups and queues user notifications.
ava-fleet
Coordinates orchestrator and worker agents with missions, task tracking, and milestone-based reporting.
ava-qa-inspection
Inspect rendered frontend pages for visual and structural defects using deterministic DOM probes.
ava-guide
Explains the ava CLI for operating clusters, agents, MCP servers, packages, presets, and schedules.
ava-corp
Organizes an Ava agent fleet into CEO, project lead, personal service, and shared infrastructure roles.
sweeper
Reconciles a repository's technical-debt tracker and lands updates as a pull request.
sms
Extract SMS verification codes and recent messages from macOS Messages database.
ava-self-evolution
Mines real agent run traces to detect skill and plugin regressions and propose fixes.
Frequently Asked Questions About Zhiyuan Zhang
FAQPage SchemaWhat tasks can I accomplish with Zhiyuan Zhang's Ava skill manifest?▼
The skills cover shipping code changes through worktrees and merge queues, adding and rolling back database migrations, deploying and operating Ava clusters, reviewing PRs adversarially, measuring McCabe complexity, transcribing audio, reading Gmail and SMS codes, ingesting RSS and YouTube sources, and running cited deep research.
Who is the intended audience for these Ava skills?▼
The manifest targets engineers and researchers operating the Ava agent platform: contributors changing kernel or plugin code, operators triaging production clusters, reviewers auditing PRs for fake-green defects, and ML researchers who need reproducible experiments, verified citations, and claim-evidence alignment.
How do the Ava cluster deployment and operations skills work in practice?▼
The deploy-ava-cluster skill installs clusters and enrolls gateway or agent-runner machines, including single-box, split, WSL, and China-mirror setups. The ops and operating-ava-cluster skills then handle lifecycle, updates, releases, and incident triage across disk, memory, connectivity, and delivery failures.
What prerequisites do these skills assume?▼
Skills assume an Ava runtime environment: the Ava CLI, git worktrees, pytest, ruff, pyright, vitest, Postgres for migrations, and observability backends like Loki, Tempo, and Grafana. Some skills require macOS Messages access, Telegram Bot credentials, or logged-in browser sessions for ChatGPT, Gemini, and Perplexity.
How do the research skills ensure trustworthy results?▼
Skills like falsifiability, reproducibility, verify, and claim-evidence-alignment enforce sealed test sets, locked environments, leakage checks, citation tracing, retraction checks, and p-hacking review. Every reported number must trace to a run, and evaluations audit agent traces for cheating without relying on ground truth.